Automobile part quality monitoring system based on big data

Abstract

The invention provides an automobile part quality monitoring system based on big data, and relates to the technical field of automobile part monitoring, and the system comprises a model training module for training a fault discrimination model, an automobile driving data collection module for collecting and restoring the running state data of an automobile, and a fault discrimination module. The power system detailed analysis module is used for carrying out real-time fault judgment and fault pre-judgment, the power system detailed analysis module is used for analyzing power system fault parts, the transmission system detailed analysis module is used for analyzing transmission system fault parts, the brake system detailed analysis module is used for analyzing brake system fault parts, and the alarm module is used for giving an alarm and early warning to the outside. According to the method, the faults are judged in real time, meanwhile, the faults are pre-judged according to the periodic data, and perspectiveness and timeliness are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile parts monitoring, and in particular to an automobile parts quality monitoring system based on big data. Background Art

[0002] With the rapid development and popularization of automobile technology, the structure of vehicles has become increasingly complex. The operating status of core components such as the engine, transmission system, and braking system has a direct impact on the performance and safety of the vehicle. However, most of the automobile fault diagnosis systems on the market are based on simple analysis of a single data source, which cannot achieve comprehensive fault identification under complex working conditions, and it is difficult to effectively warn of potential problems. The traditional way of handling automobile faults mainly relies on manual maintenance, which has the problems of low efficiency and insufficient accuracy. Especially for problems that cannot be discovered in time, it may cause the vehicle to run with problems for a long time, and even cause safety accidents. In addition, the existing technology has a weak ability to correct interference factors of operating data in complex environments, and it is difficult to ensure the reliability and accuracy of fault diagnosis results. Therefore, how to efficiently collect automobile operation data, extract features, and combine intelligent algorithms to achieve comprehensive fault identification and warning has become an important direction for the development of automobile intelligence.

[0003] In the prior art, publication number CN202311281903.7 discloses an automobile parts quality monitoring system based on big data, which includes the following steps: Step 1: Collect parameters of each welding point of automobile frame parts, and analyze welding problems occurring at each welding point during welding according to the parameters; Step 2: Use high-precision sensors to measure the welding area and calculate the area of ​​unwelded parts caused by missing welds; Step 3: The monitoring system improves the clarity of the welding appearance image based on the existing monitoring angle, monitors welding problems occurring at welding points and influencing parameters; Step 4: Monitor and sample the frame welding process, analyze the parameters that frequently affect welding based on the repair welding time of each welding point where problems occur, and optimize and adjust them; Step 5: The system visualizes the analysis results and solutions of the welding problems and generates a monitoring report.

[0004] Although the above information disclosed in the background technology section realizes the monitoring of the quality of automobile parts, the monitoring scope is small, and it is impossible to continuously monitor the automobiles put into use, and the monitoring is not forward-looking. Summary of the invention

[0005] The purpose of the present invention is to provide an automobile parts quality monitoring system based on big data to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The automobile parts quality monitoring system based on big data includes:

[0008] Model training module: Obtain vehicle fault types and vehicle operation data through the maintenance station, extract features from the vehicle operation data, obtain the maximum frequency domain component features, zero crossing rate features, acceleration features and maximum braking force features, train them in the neural network model, and obtain the fault discrimination model;

[0009] Vehicle driving data collection module: collects vehicle operation data and environmental data in real time, corrects the part of the vehicle operation data affected by the environmental data, restores the data generated by the real-time operation of the vehicle itself, and forms a driving data set;

[0010] Fault identification module: The driving data is processed in two directions. The first direction is to directly extract features and input them into the fault identification model to obtain the type of vehicle fault. The second direction is to form a future driving feature set through the change trend of the driving data set in the recent period, and input the future driving feature set into the fault identification model to obtain the type of vehicle fault.

[0011] Power system detailed analysis module: By collecting engine operating data, scoring each engine component, setting a scoring threshold, calibrating components that exceed the scoring threshold, and judging the normal working time value of the component through the time distribution model, and at the same time feeding back the normal working time value;

[0012] Transmission system detailed analysis module: By collecting transmission system operating data, the health of each transmission system component is evaluated, the health threshold is set, and the components below the health threshold are calibrated;

[0013] Braking system detailed analysis module: By judging the health of brake pads and brake lines, setting health thresholds, and calibrating components below the health thresholds;

[0014] Alarm module: For the vehicle fault type output by the fault discrimination model through feature extraction of the driving data set, an alarm is issued for the calibrated parts. For the future driving feature set generated by two consecutive cycles that outputs the same vehicle fault type as the fault discrimination model, an early warning is issued for the calibrated parts.

[0015] Furthermore, the vehicle fault type and vehicle operation data are obtained through the maintenance station, the vehicle fault type includes power system, transmission system, braking system, electrical system and no fault, the vehicle operation data includes vehicle type, vibration data of vehicle operation, sound data, power data and braking data, the vehicle operation data is sorted into an initial data set, the vehicle fault type is added to the initial data set as a label, the initial data set is feature extracted, the vehicle fault type and the vehicle type are uniquely encoded, the total cycle length is from 0:00 to 24:00 today, n sampling points are evenly delineated within the total cycle length, the vibration data is Fourier transformed to extract the maximum frequency domain component feature, and the formula is as follows:

[0016]

[0017] Among them, X(f) is the frequency domain signal, x t is the value of the tth sampling point in the time domain signal, T is the total time length, and f is the frequency;

[0018] Extract the zero crossing rate feature from the sound data based on the following formula:

[0019]

[0020] Where ZCR is the zero crossing rate, s t is the sound value of the t-th sampling point, It is an indicator function. When the condition in its brackets is met, it outputs 1, otherwise it outputs 0;

[0021] The acceleration characteristics in the dynamic data are extracted based on the following formula:

[0022]

[0023] Where a(t) is the acceleration at time t, v(t+1) and v(t) are the speeds at sampling points t+1 and t respectively, Δt is the time interval, and Y(t) is the engine fuel injection amount;

[0024] The maximum braking force feature in the braking data is extracted based on the following formula:

[0025] F brake =m·a max ·T -1

[0026] Among them, F brake is the maximum braking force, m is the mass of the vehicle, a max is the maximum deceleration, T is the brake pedal depression amount;

[0027] The maximum frequency domain component feature, zero crossing rate feature, acceleration feature and maximum braking force feature are normalized to form a feature data set, which is then input into a neural network model for training. The output result is the vehicle fault type, and the trained model is calibrated as a fault discrimination model.

[0028] Furthermore, the fault discrimination model is deployed on the car, and the car type is identified. The car collects the car operation data in real time, and collects the environmental data in real time through the on-board sensor. The environmental data includes road vibration data, environmental sound data and terrain data. The road vibration data is the data generated by the road surface when the car is driving on the road, the sound data is the sound generated by the environment around the car, and the terrain data is the impact of the slope and the road surface on the vehicle speed. The car operation data is corrected by the environmental data, and the correction logic is:

[0029] By identifying and removing road vibration data from vibration data, automobile vibration data is formed. By identifying and removing environmental sound data from sound data, automobile sound data is formed. The deceleration and acceleration at the time of the most recent inspection are obtained, and the deceleration and acceleration of the car during driving are collected in real time, and compared using the following formula:

[0030]

[0031] in, is the terrain ratio, a s is the acceleration of the car while driving, a x is the acceleration during maintenance, I is the amount of fuel supplied to the engine, aj s is the deceleration of the car during driving, aj x is the deceleration during vehicle maintenance, T is the amount of brake pedal depression;

[0032] The power data and the braking data are multiplied by coefficients to obtain vehicle power data and vehicle braking data;

[0033] The corrected car vibration data, car sound data, car power data and car braking data are sorted into a driving data set, and the car types are added to the driving data set according to the unique hot encoding.

[0034] Furthermore, the driving data set is processed in two directions at the same time. In one direction, the driving data set is subjected to feature extraction and input into a fault discrimination model to determine in real time whether a vehicle fault occurs. When the output vehicle fault type is not fault-free, an alarm is issued for the calibrated parts. When the output vehicle fault type is fault-free, no alarm is issued.

[0035] In another direction, the driving data sets of the latest k cycles are compared with each other on a daily basis, and the maximum frequency domain component features, zero crossing rate features, acceleration features, and maximum braking force features of each cycle are extracted respectively. The change trends of the latest k cycles are analyzed respectively, where the change trends include the maximum frequency domain component feature change rate, zero crossing feature change rate, acceleration feature change rate, and maximum braking force feature change rate. The maximum frequency domain component feature change rate is identified based on the following formula:

[0036]

[0037] Among them, R total (k) is the rate of change within k cycles, F is the set of all maximum frequency domain components within k cycles, X i (f) is the amplitude of frequency f in the ith period;

[0038] Identify the rate of change of the zero-crossing rate feature, based on the following formula:

[0039]

[0040] Among them, ZCRR is the zero crossing rate change rate, ZCR i is the zero-crossing rate characteristic value of the i-th period, i is the search variable, i∈N, 1≤i≤k-1;

[0041] Identify the acceleration characteristic change rate based on the following formula:

[0042]

[0043] Among them, A is the acceleration characteristic change rate, a i is the average acceleration of the i-th period, i is the search variable, i∈N, 1≤i≤k-1;

[0044] The maximum braking force characteristic change rate is identified based on the following formula:

[0045]

[0046] Among them, F RR is the maximum braking force characteristic change rate, F brake- i is the maximum braking force characteristic of the i-th cycle, i is the retrieval variable, i∈N, 1≤i≤k-1;

[0047] Based on the periodic data of today, feature extraction is performed. According to the change trend of the driving data set of the k periods before today, each feature is multiplied by the corresponding change rate to generate the driving feature set of the next 15 periods. The driving feature sets of these 15 periods are input into the fault discrimination model respectively. When one of the future driving feature sets generates a vehicle fault type that is not fault-free, it is determined that the future driving feature set generated in this period input into the fault discrimination model has generated a vehicle fault type that is not fault-free.

[0048] Furthermore, when the fault discrimination model determines that the fault is caused by the power system, the power system is analyzed in detail to obtain engine operating data, including vibration data, pressure data, temperature data, total operating time, full-load operating time, and the most recent maintenance score. Parts for which engine operating data can be directly obtained are numbered and scored using the following formula:

[0049] H i =w1·D i +w2·R i +w3·S i +w4·F i +w5·G i +w6·H i +w7·I i -1

[0050] Among them, H i is the score of the component numbered i, D i is the maximum amplitude of the component numbered i, R i is the total working time of the component numbered i, S i is the maximum pressure of the component numbered i, F i is the maximum temperature of the component numbered i, G i is the total working time of the component numbered i, H i is the full load working time of the component numbered i, I i is the most recent maintenance score of the component numbered i, i is the component number search variable, i∈N, 1≤i≤n;

[0051] w1, w2, w3, w4, w5, w6, w7 are weight coefficients satisfying:

[0052] w1≥w2≥w3≥w4≥w5≥w6≥w7

[0053] w1+w2+w3+w4+w5+w6+w7=1

[0054] For the scores of parts that cannot directly obtain working condition data, the scores of parts that can directly obtain working condition data are weighted according to the matching relationship. The formula is as follows:

[0055]

[0056] Among them, G j Score the component numbered j, H i is the score of the i-th component that has a matching relationship with the component numbered j, τ is the weighting coefficient, i is the component number retrieval variable, i∈N, 1≤i≤u;

[0057] Set the scoring threshold of each component. When the corresponding component score exceeds the scoring threshold, calibrate the component and use the failure time distribution model to determine the normal working value of the component. The formula is as follows:

[0058] R(t)=e -(λt)β

[0059] Where R(t) is the probability that the component can continue to work within time t, and λ and β are parameters;

[0060] Set a probability threshold, obtain the normal working time value of the component, and feedback the normal working time value of the component.

[0061] Furthermore, when the fault discrimination model determines that the fault is a transmission system, the transmission system is analyzed in detail to obtain the transmission system operating data, which includes vibration data, temperature data, torque data, driving mode and duration data, working years, full-load working time, historical fault times and historical maintenance times of each component. The health assessment of each component is performed based on the following formula:

[0062] H=w1·H state +w2·H age +w3·H history +w4·H maintenance

[0063] Among them, H is the health assessment value of this component, H state is the reference value of real-time monitoring data, H age is the annual reference value, H history is the historical operating condition reference value, H maintenance is the reference value of maintenance records; w1, w2, w3, w4 are weight coefficients, where:

[0064] w1+w2+w3+w4=1

[0065] The formula for the reference value of real-time monitoring data is as follows:

[0066]

[0067] Among them, σ vibration is the standard deviation of vibration amplitude, σ temperature is the temperature standard deviation, σ torque is the torque standard deviation, α is the weight factor;

[0068] The formula for the annual reference value is as follows:

[0069] H age =e -β·t

[0070] Where t is the service life of the component and β is the attenuation coefficient;

[0071] The historical operating condition reference value is based on the following formula:

[0072]

[0073] Among them, f i is the influence parameter of the ith driving mode on the component, where the driving mode is the preset driving mode of the car adopted by the driver, t i is the duration of the driving mode, γ is the influence coefficient;

[0074] The maintenance record reference value is based on the following formula:

[0075] H maintenance =1-δ·(N failure +N repair )

[0076] Among them, N failure is the number of failures, N repair is the number of maintenance times, δ is the influence coefficient;

[0077] For components that can directly obtain working condition data, their health valuation is directly evaluated. For components that cannot directly obtain working condition data, the health valuation is obtained by weighting the health valuation of components that can directly obtain working condition data according to the matching relationship. The formula is as follows:

[0078]

[0079] Among them, G j is the health evaluation of component number j, H i is the health valuation of the i-th component that has a matching relationship with the component numbered j, τ is the weighting coefficient, i is the component number retrieval variable, i∈N, 1≤i≤u;

[0080] Set the health threshold of each component. When the corresponding component health valuation is lower than the health threshold, the component is marked as a problem component.

[0081] Furthermore, when the fault discrimination model determines that the fault is in the brake system, the brake system is analyzed in detail, and the health of the brake pads and brake lines is judged respectively. The logic for judging the health of the brake pads is as follows:

[0082] Get the current brake pad thickness, initial brake pad thickness and service life to perform brake pad health assessment. The formula is as follows:

[0083]

[0084] Among them, Q pad For brake pad health, T current is the current brake pad thickness, T max is the initial thickness of the brake pad, D age is the useful life, D env is the operating factor, which is determined by the type of vehicle;

[0085] The logic for judging the health of the brake line is as follows:

[0086] Get the current management summary brake fluid pressure, maximum working pressure of the pipeline, usage time and the most recent maintenance score. The health assessment is based on the following formula:

[0087]

[0088] Among them, Q line is the health of the brake line, P current is the brake fluid pressure in the current pipeline, P max is the maximum working pressure of the pipeline, D age is the useful life, D maint Rate the most recent maintenance;

[0089] The brake pad health threshold and the brake line health threshold are set respectively. When the brake pad health or the brake line health is lower than the corresponding threshold, the brake pad or the brake line is calibrated.

[0090] Furthermore, when the fault identification model determines that the fault is related to the electrical system, the driver will manually determine the cause of the fault and perform preliminary repairs. When the driver is unable to manually determine the fault or cannot repair it, the vehicle will be sent to a professional organization for determination and repair.

[0091] Furthermore, when the fault discrimination model determines that there is no fault, the detailed analysis of the power system, transmission system and braking system does not work. When the driving data acquired in the current period is subjected to feature extraction and input into the fault discrimination model to obtain a vehicle fault type that is not fault-free, an alarm is given to the calibrated parts. When a vehicle fault type that is not fault-free is obtained by inputting the fault discrimination model with a future driving feature set, no immediate warning is given. When the future driving feature set generated by two consecutive periods and input into the fault discrimination model both produce the same vehicle fault type that is not fault-free, an alarm is given to the calibrated parts.

[0092] Compared with the prior art, the present invention has the following beneficial effects:

[0093] The present invention integrates multiple data to train a fault discrimination model, and uses real-time monitoring data to judge current faults and possible future faults. It is timely and forward-looking, and through detailed analysis of each system, it accurately judges faulty components and ensures the normal operation of the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Figure 1 It is a schematic diagram of the overall structure of the present invention. DETAILED DESCRIPTION

[0095] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0096] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0097] Example:

[0098] See also Figure 1 , the present invention provides a technical solution:

[0099] The automobile parts quality monitoring system based on big data includes:

[0100] Model training module: Obtain vehicle fault types and vehicle operation data through the maintenance station, extract features from the vehicle operation data, obtain the maximum frequency domain component features, zero crossing rate features, acceleration features and maximum braking force features, train them in the neural network model, and obtain the fault discrimination model;

[0101] Vehicle driving data collection module: collects vehicle operation data and environmental data in real time, corrects the part of the vehicle operation data affected by the environmental data, restores the data generated by the real-time operation of the vehicle itself, and forms a driving data set;

[0102] Fault identification module: The driving data is processed in two directions. The first direction is to directly extract features and input them into the fault identification model to obtain the type of vehicle fault. The second direction is to form a future driving feature set through the change trend of the driving data set in the recent period, and input the future driving feature set into the fault identification model to obtain the type of vehicle fault.

[0103] Power system detailed analysis module: By collecting engine operating data, scoring each engine component, setting a scoring threshold, calibrating components that exceed the scoring threshold, and judging the normal working time value of the component through the time distribution model, and at the same time feeding back the normal working time value;

[0104] Transmission system detailed analysis module: By collecting transmission system operating data, the health of each transmission system component is evaluated, the health threshold is set, and the components below the health threshold are calibrated;

[0105] Braking system detailed analysis module: By judging the health of brake pads and brake lines, setting health thresholds, and calibrating components below the health thresholds;

[0106] Alarm module: For the vehicle fault type output by the fault discrimination model through feature extraction of the driving data set, an alarm is issued for the calibrated parts. For the future driving feature set generated by two consecutive cycles that outputs the same vehicle fault type as the fault discrimination model, an early warning is issued for the calibrated parts.

[0107] As a preferred embodiment, the automobile fault type and automobile operation data are obtained through the maintenance station, the automobile fault type includes power system, transmission system, braking system, electrical system and no fault, the automobile operation data includes vehicle type, vibration data of vehicle operation, sound data, power data and braking data, the automobile operation data is sorted into an initial data set, the automobile fault type is added to the initial data set as a label, the initial data set is subjected to feature extraction, the automobile fault type and the automobile type are respectively subjected to unique hot encoding, each automobile fault type is three-dimensionally encoded and complementary, the automobile type is a classification of all automobile types on the market, each automobile type classification is five-dimensionally encoded and different from each other, the total cycle length is from 0:00 to 24:00 today, n sampling points are evenly delineated within the total cycle length, the vibration data is subjected to Fourier transform to extract the maximum frequency domain component feature, and the formula based on it is as follows:

[0108]

[0109] Among them, X(f) is the frequency domain signal, x t is the value of the tth sampling point in the time domain signal, T is the total time length, and f is the frequency;

[0110] The zero crossing rate feature in the sound data is extracted based on the following formula:

[0111]

[0112] Where ZCR is the zero crossing rate, s t is the sound value of the t-th sampling point, It is an indicator function. When the condition in the brackets is met, it outputs 1, otherwise it outputs 0;

[0113] The acceleration characteristics in the dynamic data are extracted based on the following formula:

[0114]

[0115] Where a(t) is the acceleration at time t, v(t+1) and v(t) are the speeds at sampling points t+1 and t respectively, Δt is the time interval, and Y(t) is the engine fuel injection amount;

[0116] The amount of fuel injected into the engine determines the car's ability to accelerate. The greater the amount of fuel injected into the engine, the smaller the car's acceleration characteristics.

[0117] The maximum braking force feature in the braking data is extracted based on the following formula:

[0118] F brake =m·a max ·T -1

[0119] Among them, F brake is the maximum braking force, m is the mass of the vehicle, a max is the maximum deceleration, T is the brake pedal depression amount;

[0120] The braking capacity of a car is determined by the amount the brake pedal is depressed; the greater the amount of depression, the smaller the braking force.

[0121] The maximum frequency domain component feature, zero crossing rate feature, acceleration feature and maximum braking force feature are normalized to form a feature data set, which is then input into a neural network model for training. The output result is the vehicle fault type, and the trained model is calibrated as a fault discrimination model.

[0122] The vibration data, sound data, power data and braking data of all types of vehicles can increase the adaptability of the fault discrimination model. In the process of feature extraction, continuous variables are divided into discrete variables by defining sampling points within the total cycle length, thereby reducing the computational complexity of model training.

[0123] As a preferred embodiment, the fault discrimination model is deployed on the car, and the car type is identified. The car collects the car operation data in real time, and collects the environmental data in real time through the on-board sensor. The environmental data includes road vibration data, environmental sound data and terrain data. The road vibration data is the data generated by the road surface when the car is driving on the road. The sound data is the sound generated by the environment around the car. The terrain data is the impact of the slope and the road surface on the vehicle speed. The car operation data is corrected by the environmental data, and the correction logic is:

[0124] By identifying and removing road vibration data from vibration data, automobile vibration data is formed. By identifying and removing environmental sound data from sound data, automobile sound data is formed. The deceleration and acceleration at the time of the most recent inspection are obtained, and the deceleration and acceleration of the car during driving are collected in real time, and compared using the following formula:

[0125]

[0126] in, is the terrain ratio, a s is the acceleration of the car while driving, a x is the acceleration during maintenance, I is the amount of fuel supplied to the engine, aj s is the deceleration of the car during driving, aj x is the deceleration during vehicle maintenance, T is the amount of brake pedal depression;

[0127] By averaging the changes in acceleration and deceleration of the car on the slope, the terrain ratio of the car's location can be determined, providing a basis for obtaining accurate vehicle power data and vehicle braking data in the next step.

[0128] The power data and the braking data are multiplied by coefficients to obtain vehicle power data and vehicle braking data;

[0129] The corrected car vibration data, car sound data, car power data and car braking data are sorted into a driving data set, and the car types are added to the driving data set according to the unique hot encoding.

[0130] By collecting vehicle operation data and environmental data at the same time, the part of the vehicle operation data affected by the environmental data is eliminated, and the vehicle's own operation data is restored to the greatest extent. When the feature extraction is subsequently input into the fault discrimination model, the vehicle's own fault type can be more accurately judged.

[0131] As a preferred embodiment, the driving data set is processed in two directions at the same time. In one direction, the driving data set is subjected to feature extraction and input into a fault discrimination model to determine in real time whether a vehicle fault occurs. When the output vehicle fault type is not fault-free, an alarm is issued for the calibrated parts. When the output vehicle fault type is fault-free, no alarm is issued.

[0132] By extracting features from the driving data set in real time and inputting them into the fault identification model, the faults that occur can be analyzed in real time, thereby identifying sudden faults and sending alarms to remind personnel to perform timely maintenance.

[0133] In another direction, the driving data sets of the latest k cycles are compared with each other on a daily basis, and the maximum frequency domain component features, zero crossing rate features, acceleration features, and maximum braking force features of each cycle are extracted respectively. The change trends of the latest k cycles are analyzed respectively, where the change trends include the maximum frequency domain component feature change rate, zero crossing feature change rate, acceleration feature change rate, and maximum braking force feature change rate. The maximum frequency domain component feature change rate is identified based on the following formula:

[0134]

[0135] Among them, R total (k) is the rate of change within k cycles, F is the set of all maximum frequency domain components within k cycles, X i (f) is the amplitude of frequency f in the ith period;

[0136] Identify the rate of change of the zero-crossing rate feature, based on the following formula:

[0137]

[0138] Among them, ZCRR is the zero crossing rate change rate, ZCR i is the zero-crossing rate characteristic value of the i-th period, i is the search variable, i∈N, 1≤i≤k-1;

[0139] Identify the acceleration characteristic change rate based on the following formula:

[0140]

[0141] Among them, A is the acceleration characteristic change rate, a i is the average acceleration of the i-th period, i is the search variable, i∈N, 1≤i≤k-1;

[0142] The maximum braking force characteristic change rate is identified based on the following formula:

[0143]

[0144] Among them, F RR is the maximum braking force characteristic change rate, F brake-i is the maximum braking force characteristic of the i-th cycle, i is the retrieval variable, i∈N, 1≤i≤k-1;

[0145] Based on the periodic data of today, feature extraction is performed. According to the change trend of the driving data set of the k periods before today, each feature is multiplied by the corresponding change rate to generate the driving feature set of the next 15 periods. The driving feature sets of these 15 periods are input into the fault discrimination model respectively. When one of the future driving feature sets generates a vehicle fault type that is not fault-free, it is determined that the future driving feature set generated in this period input into the fault discrimination model has generated a vehicle fault type that is not fault-free.

[0146] The future driving feature set is generated through the formal data of the daily cycle. The driving feature set reflects the changes in automobile parts in the short term, thereby realizing the prediction of the operating status of the automobile in the future.

[0147] As a preferred embodiment, when the fault discrimination model determines that the result is the power system, the power system is analyzed in detail to obtain engine operating condition data, which includes vibration data, pressure data, temperature data, total working time, full-load working time and the most recent maintenance score. Parts for which engine operating condition data can be directly obtained are numbered and scored using the following formula:

[0148] H i =w1·D i +w2·R i +w3·S i +w4·F i +w5·Gi +w6·H i +w7·I i -1

[0149] Among them, H i is the score of the component numbered i, D i is the maximum amplitude of the component numbered i, R i is the total working time of the component numbered i, S i is the maximum pressure of the component numbered i, F i is the maximum temperature of the component numbered i, G i is the total working time of the component numbered i, H i is the full load working time of the component numbered i, I i is the most recent maintenance score of the component numbered i, i is the component number search variable, i∈N, 1≤i≤n;

[0150] The score is obtained by multiplying the proportion of the loss of each state of the component. The higher the last maintenance score of the component, the lower the score of the component, which means the loss of the component is smaller. The greater the other losses, the greater the score of the component. When the threshold is exceeded, the loss of the component exceeds the critical state.

[0151] w1, w2, w3, w4, w5, w6, w7 are weight coefficients satisfying:

[0152] w1≥w2≥w3≥w4≥w5≥w6≥w7

[0153] w1+w2+w3+w4+w5+w6+w7=1

[0154] For the scores of parts that cannot directly obtain working condition data, the scores of parts that can directly obtain working condition data are weighted according to the matching relationship. The formula is as follows:

[0155]

[0156] Among them, G j Score the component numbered j, H i is the score of the i-th component that has a matching relationship with the component numbered j, τ is the weighting coefficient, i is the component number retrieval variable, i∈N, 1≤i≤u;

[0157] The average loss of matching parts is used to replace the score of the component that cannot directly obtain the working condition data, which means that the greater the loss of the matching parts, the greater the loss of the part.

[0158] Set the scoring threshold of each component. When the corresponding component score exceeds the scoring threshold, calibrate the component and use the failure time distribution model to determine the normal working value of the component. The formula is as follows:

[0159]

[0160] Where R(t) is the probability that the component can continue to work within time t, and λ and β are parameters;

[0161] Set a probability threshold, obtain the normal working time value of the component, and feedback the normal working time value of the component.

[0162] By collecting engine operating condition data, each engine component can be monitored. For components that cannot directly obtain operating conditions, their working status can be indirectly reflected through the cooperating components, thereby realizing the monitoring of all engine components and providing timely warnings when faults occur.

[0163] As a preferred embodiment, when the judgment result of the fault discrimination model is the transmission system, the transmission system is analyzed in detail to obtain the transmission system working condition data, which includes the vibration data, temperature data, torque data, driving mode and duration data, working years, full load working time, historical fault times and historical maintenance times of each component, and the health assessment of each component is performed according to the following formula:

[0164] H=w1·H state +w2·H age +w3·H history +w4·H maintenance

[0165] Among them, H is the health assessment value of this component, H state is the reference value of real-time monitoring data, H age is the annual reference value, H history is the historical operating condition reference value, H maintenance is the reference value of maintenance records; w1, w2, w3, w4 are weight coefficients, where:

[0166] w1+w2+w3+w4=1

[0167] The formula for the reference value of real-time monitoring data is as follows:

[0168]

[0169] Among them, σ vibration is the standard deviation of vibration amplitude, σ temperature is the temperature standard deviation, σ torque is the torque standard deviation, α is the weight factor;

[0170] As the standard deviation of vibration amplitude increases, the standard deviation of temperature increases, and the standard deviation of torque increases, it means that the reference value of the real-time monitoring data of the component will be lower, thereby reducing the health assessment value of the component.

[0171] The formula for the annual reference value is as follows:

[0172] H age =e -β·t

[0173] Where t is the service life of the component and β is the attenuation coefficient;

[0174] As the service life increases, the wear and tear of the component increases, so the age reference value decreases, thereby reducing the health assessment value of the component.

[0175] The historical operating condition reference value is based on the following formula:

[0176]

[0177] Among them, f i is the influence parameter of the ith driving mode on the component, where the driving mode is the preset driving mode of the car adopted by the driver, t i is the duration of the driving mode, γ is the influence coefficient;

[0178] This formula represents the damage to the vehicle's components caused by driving habits. The more violent the driving mode and the longer the driving time, the lower the historical operating condition reference value will be, thereby reducing the health assessment value of the component.

[0179] The maintenance record reference value is based on the following formula:

[0180] H maintenance =1-δ·(N failure +N repair )

[0181] Among them, N failure is the number of failures, N repair is the number of maintenance times, δ is the influence coefficient;

[0182] An increase in the number of breakdowns and repairs means that the condition of the car has deteriorated, which reduces the reference value of the maintenance record and thus reduces the health assessment value of the component.

[0183] For components that can directly obtain working condition data, their health valuation is directly evaluated. For components that cannot directly obtain working condition data, the health valuation is obtained by weighting the health valuation of components that can directly obtain working condition data according to the matching relationship. The formula is as follows:

[0184]

[0185] Among them, G j is the health evaluation of component number j, H i is the health valuation of the i-th component that has a matching relationship with the component numbered j, τ is the weighting coefficient, i is the component number retrieval variable, i∈N, 1≤i≤u;

[0186] Set the health threshold of each component. When the corresponding component health valuation is lower than the health threshold, the component is marked as a problem component.

[0187] By acquiring the transmission system operating data, the health of each component of the transmission system is evaluated, thereby obtaining the actual status of each component and realizing quality monitoring of each component in the transmission system.

[0188] As a preferred embodiment, when the fault discrimination model determines that the fault is a braking system, the braking system is analyzed in detail, and the health of the brake pads and brake lines is judged respectively. The logic for judging the health of the brake pads is as follows:

[0189] Get the current brake pad thickness, initial brake pad thickness and service life to perform brake pad health assessment. The formula is as follows:

[0190]

[0191] Among them, Q pad For brake pad health, T current is the current brake pad thickness, T max is the initial thickness of the brake pad, D age is the useful life, D env is the operating factor, which is determined by the type of vehicle;

[0192] The smaller the ratio of the current brake pad thickness to the initial brake pad thickness, the greater the wear of the brake pad, and the longer the service life, the lower the health of the brake pad.

[0193] The logic for judging the health of the brake line is as follows:

[0194] Get the current management summary brake fluid pressure, maximum working pressure of the pipeline, usage time and the most recent maintenance score. The health assessment is based on the following formula:

[0195]

[0196] Among them, Q line is the health of the brake line, P current is the brake fluid pressure in the current pipeline, P maxis the maximum working pressure of the pipeline, D age is the useful life, D maint Rate the most recent maintenance;

[0197] The lower the ratio of the current brake fluid pressure in the pipeline to the maximum working pressure of the pipeline, the greater the problem with the pipeline, and the older the service life, the lower the health of the brake pipeline.

[0198] The brake pad health threshold and the brake line health threshold are set respectively. When the brake pad health or the brake line health is lower than the corresponding threshold, the brake pad or the brake line is calibrated.

[0199] By obtaining the main components in the braking system for identification, the fault identification process is simplified and the identification efficiency is improved.

[0200] As a preferred embodiment, when the fault identification model determines that the fault is an electrical system, the driver manually determines the cause of the fault and performs preliminary repairs. When the driver is unable to manually determine the fault or cannot repair it, the vehicle is sent to a professional organization for determination and repair.

[0201] As a preferred embodiment, when the fault discrimination model determines that the fault is not fault-free, the detailed analysis of the power system, transmission system and braking system does not work. When the driving data acquired in the current period is subjected to feature extraction and input into the fault discrimination model to obtain a vehicle fault type that is not fault-free, an alarm is given to the calibrated parts. When the future driving feature set is input into the fault discrimination model to obtain a vehicle fault type that is not fault-free, no immediate warning is given. When the future driving feature set generated in two consecutive periods that is input into the fault discrimination model both produces the same vehicle fault type that is not fault-free, an alarm is given to the calibrated parts.

[0202] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0203] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0204] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0205] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. The automobile parts quality monitoring system based on big data is characterized by: include: Model training module: Obtain vehicle fault types and vehicle operation data through the maintenance station, extract features from the vehicle operation data, obtain the maximum frequency domain component features, zero crossing rate features, acceleration features and maximum braking force features, train them in the neural network model, and obtain the fault discrimination model; Vehicle driving data collection module: collects vehicle operation data and environmental data in real time, corrects the part of the vehicle operation data affected by the environmental data, restores the data generated by the real-time operation of the vehicle itself, and forms a driving data set; Fault identification module: The driving data is processed in two directions. The first direction is to directly extract features and input them into the fault identification model to obtain the type of vehicle fault. The second direction is to form a future driving feature set through the change trend of the driving data set in the recent period, and input the future driving feature set into the fault identification model to obtain the type of vehicle fault. Power system detailed analysis module: By collecting engine operating data, scoring each engine component, setting a scoring threshold, calibrating components that exceed the scoring threshold, and judging the normal working time value of the component through the time distribution model, and at the same time feeding back the normal working time value; Transmission system detailed analysis module: By collecting transmission system operating data, the health of each transmission system component is evaluated, the health threshold is set, and the components below the health threshold are calibrated; Braking system detailed analysis module: By judging the health of brake pads and brake lines, setting health thresholds, and calibrating components below the health thresholds; Alarm module: For the vehicle fault type output by the fault discrimination model through feature extraction of the driving data set, an alarm is issued for the calibrated parts. For the future driving feature set generated by two consecutive cycles that outputs the same vehicle fault type as the fault discrimination model, an early warning is issued for the calibrated parts.

2. The automobile parts quality monitoring system based on big data according to claim 1 is characterized by: Obtain automobile fault types and automobile operation data through the maintenance station. The automobile fault types include power system, transmission system, braking system, electrical system and no fault. The automobile operation data includes vehicle type, vibration data of vehicle operation, sound data, power data and braking data. The automobile operation data is sorted into an initial data set, and the automobile fault type is added to the initial data set as a label. The initial data set is feature extracted, and the automobile fault type and automobile type are uniquely encoded. The total cycle length is from 0:00 to 24:00 today, and n sampling points are evenly delineated within the total cycle length. The vibration data is Fourier transformed to extract the maximum frequency domain component feature. The formula is as follows: Among them, X(f) is the frequency domain signal, x t is the value of the tth sampling point in the time domain signal, T is the total time length, and f is the frequency; Extract the zero crossing rate feature from the sound data based on the following formula: Among them, ZCR is the zero crossing rate, st is the sound value of the t-th sampling point, It is an indicator function. When the condition in its brackets is met, it outputs 1, otherwise it outputs 0; The acceleration characteristics in the dynamic data are extracted based on the following formula: Where a(t) is the acceleration at time t, v(t+1) and v(t) are the speeds at sampling points t+1 and t respectively, Δt is the time interval, and Y(t) is the engine fuel injection amount; The maximum braking force feature in the braking data is extracted based on the following formula: F brake =m·a max ·T -1 Among them, F brake is the maximum braking force, m is the mass of the vehicle, a max is the maximum deceleration, T is the brake pedal depression amount; The maximum frequency domain component feature, zero crossing rate feature, acceleration feature and maximum braking force feature are normalized to form a feature data set, which is then input into a neural network model for training. The output result is the vehicle fault type, and the trained model is calibrated as a fault discrimination model.

3. The automobile parts quality monitoring system based on big data according to claim 2 is characterized in that: The fault discrimination model is deployed on the car and the car type is identified. The car collects the car operation data in real time and collects the environmental data in real time through the on-board sensor. The environmental data includes road vibration data, environmental sound data and terrain data. The road vibration data is the data generated by the road surface when the car is driving on the road. The sound data is the sound generated by the environment around the car. The terrain data is the impact of the slope and road surface on the vehicle speed. The car operation data is corrected by the environmental data. The correction logic is: By identifying and removing road vibration data from vibration data, automobile vibration data is formed. By identifying and removing environmental sound data from sound data, automobile sound data is formed. The deceleration and acceleration at the time of the most recent inspection are obtained, and the deceleration and acceleration of the car during driving are collected in real time, and compared using the following formula: in, is the terrain scale, a s is the acceleration of the car while driving, a x is the acceleration during maintenance, I is the amount of fuel supplied to the engine, aj s is the deceleration of the car during driving, aj x is the deceleration during vehicle maintenance, T is the amount of brake pedal depression; The power data and the braking data are multiplied by coefficients to obtain vehicle power data and vehicle braking data; The corrected car vibration data, car sound data, car power data and car braking data are sorted into a driving data set, and the car types are added to the driving data set according to the unique hot encoding.

4. The automobile parts quality monitoring system based on big data according to claim 3 is characterized in that: The driving data set is processed in two directions at the same time. In one direction, the driving data set is feature extracted and input into the fault discrimination model to determine in real time whether the vehicle has a fault. When the output vehicle fault type is not fault-free, the calibrated parts are alarmed. When the output vehicle fault type is fault-free, no alarm is given. In another direction, the driving data sets of the latest k cycles are compared with each other on a daily basis, and the maximum frequency domain component features, zero crossing rate features, acceleration features, and maximum braking force features of each cycle are extracted respectively. The change trends of the latest k cycles are analyzed respectively, where the change trends include the maximum frequency domain component feature change rate, zero crossing feature change rate, acceleration feature change rate, and maximum braking force feature change rate. The maximum frequency domain component feature change rate is identified based on the following formula: Among them, R total (k) is the rate of change within k cycles, F is the set of all maximum frequency domain components within k cycles, X i (f) is the amplitude of frequency f in the ith period; Identify the rate of change of the zero-crossing rate feature, based on the following formula: Among them, ZCRR is the zero crossing rate change rate, ZCR i is the zero-crossing rate characteristic value of the i-th period, i is the search variable, i∈N, 1≤i≤k-1; Identify the acceleration characteristic change rate based on the following formula: Among them, A is the acceleration characteristic change rate, a i is the average acceleration of the i-th cycle, i is the search variable, i∈N, 1≤i≤k-1; The maximum braking force characteristic change rate is identified based on the following formula: Among them, F RR is the maximum braking force characteristic change rate, F brake-i is the maximum braking force characteristic of the i-th cycle, i is the retrieval variable, i∈N, 1≤i≤k-1; Based on the periodic data of today, feature extraction is performed. According to the change trend of the driving data set of the k periods before today, each feature is multiplied by the corresponding change rate to generate the driving feature set of the next 15 periods. The driving feature sets of these 15 periods are input into the fault discrimination model respectively. When one of the future driving feature sets generates a vehicle fault type that is not fault-free, it is determined that the future driving feature set generated in this period input into the fault discrimination model has generated a vehicle fault type that is not fault-free.

5. The automobile parts quality monitoring system based on big data according to claim 4 is characterized in that: When the fault discrimination model determines that the fault is caused by the power system, the power system is analyzed in detail to obtain engine operating data, including vibration data, pressure data, temperature data, total operating time, full-load operating time, and the most recent maintenance score. Parts that can directly obtain engine operating data are numbered and scored using the following formula: <h2 style=";text-align:left;direction:ltr">H<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> =w1·D<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> +w2·R<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> +w3·S<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> +w4·F<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> +w5·G<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> +w6·H<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> +w7·I<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> -1 Among them, H i is the score of the component numbered i, D i is the maximum amplitude of the component numbered i, R i is the total working time of the component numbered i, S i is the maximum pressure of the component numbered i, F i is the maximum temperature of the component numbered i, G i is the total working time of the component numbered i, H i is the full load working time of the component numbered i, I i is the most recent maintenance score of the component numbered i, i is the component number search variable, i∈N, 1≤i≤n; w1, w2, w3, w4, w5, w6, w7 are weight coefficients satisfying: w1≥w2≥w3≥w4≥w5≥w6≥w7 w1+w2+w3+w4+w5+w6+w7=1 For the scores of parts that cannot directly obtain working condition data, the scores of parts that can directly obtain working condition data are weighted according to the matching relationship. The formula is as follows: Among them, G j Score the component numbered j, H i is the score of the i-th component that has a matching relationship with the component numbered j, τ is the weighting coefficient, i is the component number retrieval variable, i∈N, 1≤i≤u; Set the scoring threshold of each component. When the corresponding component score exceeds the scoring threshold, calibrate the component and use the failure time distribution model to determine the normal working value of the component. The formula is as follows: Where R(t) is the probability that the component can continue to work within time t, and λ and β are parameters; Set a probability threshold, obtain the normal working time value of the component, and feedback the normal working time value of the component.

6. The automobile parts quality monitoring system based on big data according to claim 5 is characterized in that: When the fault discrimination model determines that the fault discrimination model is a transmission system, the transmission system is analyzed in detail to obtain the transmission system operating data, which includes vibration data, temperature data, torque data, driving mode and duration data, working years, full-load working time, historical fault times and historical maintenance times of each component. The health assessment of each component is performed based on the following formula: H=w1·H state +w2·H age +w3·H history +w4·H maintenance Among them, H is the health assessment value of this component, H state is the reference value of real-time monitoring data, H age is the annual reference value, H history is the historical operating condition reference value, H maintenance is the reference value of maintenance records; w1, w2, w3, w4 are weight coefficients, where: w1+w2+w3+w4=1 The formula for the reference value of real-time monitoring data is as follows: Among them, σ vibration is the standard deviation of vibration amplitude, σ temperature is the temperature standard deviation, σ torque is the torque standard deviation, α is the weight factor; The formula for the annual reference value is as follows: H age =e -β·t Where t is the service life of the component and β is the attenuation coefficient; The historical operating condition reference value is based on the following formula: Among them, f i is the influence parameter of the ith driving mode on the component, where the driving mode is the preset driving mode of the car adopted by the driver, t i is the duration of the driving mode, γ is the influence coefficient; The maintenance record reference value is based on the following formula: H maintenance =1-δ·(N failure +N repair ) Among them, N failure is the number of failures, N repair is the number of maintenance times, δ is the influence coefficient; For components that can directly obtain working condition data, their health valuation is directly evaluated. For components that cannot directly obtain working condition data, the health valuation is obtained by weighting the health valuation of components that can directly obtain working condition data according to the matching relationship. The formula is as follows: Among them, G j is the health evaluation of component number j, H i is the health valuation of the i-th component that has a matching relationship with the component numbered j, τ is the weighting coefficient, i is the component number retrieval variable, i∈N, 1≤i≤u; Set the health threshold of each component. When the corresponding component health valuation is lower than the health threshold, the component is marked as a problem component.

7. The automobile parts quality monitoring system based on big data according to claim 6 is characterized by: When the fault discrimination model determines that the fault is in the braking system, the braking system is analyzed in detail, and the health of the brake pads and brake lines is judged respectively. The logic for judging the health of the brake pads is as follows: Get the current brake pad thickness, initial brake pad thickness and service life to perform brake pad health assessment. The formula is as follows: Among them, Q pad For brake pad health, T current is the current brake pad thickness, T max is the initial thickness of the brake pad, D age is the useful life, D env is the operating factor, which is determined by the type of vehicle; The logic for judging the health of the brake line is as follows: Get the current management summary brake fluid pressure, maximum working pressure of the pipeline, usage time and the most recent maintenance score. The health assessment is based on the following formula: Among them, Q line is the health of the brake line, P current is the brake fluid pressure in the current pipeline, P max is the maximum working pressure of the pipeline, D age is the useful life, D maint Rate the most recent maintenance; The brake pad health threshold and the brake line health threshold are set respectively. When the brake pad health or the brake line health is lower than the corresponding threshold, the brake pad or the brake line is calibrated.

8. The automobile parts quality monitoring system based on big data according to claim 7 is characterized in that: When the fault identification model determines that the fault is in the electrical system, the driver will manually determine the cause of the fault and perform preliminary repairs. When the driver is unable to manually determine the fault or cannot repair it, the vehicle will be sent to a professional organization for determination and repair.

9. The automobile parts quality monitoring system based on big data according to claim 8 is characterized in that: When the fault discrimination model determines that there is no fault, the detailed analysis of the power system, transmission system and braking system does not work. When the driving data obtained in the current cycle is extracted and input into the fault discrimination model to obtain a vehicle fault type that is not fault-free, the calibrated parts are alarmed. When the future driving feature set is input into the fault discrimination model to obtain a vehicle fault type that is not fault-free, no immediate warning is given. When the future driving feature set generated by two consecutive cycles input into the fault discrimination model both produces the same vehicle fault type that is not fault-free, a warning is given to the calibrated parts.

Images